![[2412.18362] Point-DeepONet: Predicting Nonlinear Fields on Non-Parametric Geometries under Variable Load Conditions](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2412.18362] Point-DeepONet: Predicting Nonlinear Fields on Non-Parametric Geometries under Variable Load Conditions
Summary
Point-DeepONet introduces a novel approach for predicting nonlinear fields in engineering, leveraging deep learning to enhance efficiency in structural analyses under variable load conditions.
Why It Matters
This research addresses the limitations of traditional finite element methods in engineering, offering a faster and more accurate alternative for structural analysis. By integrating PointNet with DeepONet, it provides significant improvements in computational efficiency and accuracy, which is crucial for real-time applications in engineering design and optimization.
Key Takeaways
- Point-DeepONet combines PointNet and DeepONet for efficient predictions.
- Achieves up to 400 times faster predictions compared to traditional methods.
- Demonstrates high accuracy with R² values of 0.987 for displacement and 0.923 for stress.
- Validates generalization capabilities through experiments with unseen load conditions.
- Offers potential for rapid structural analyses in complex engineering workflows.
Computer Science > Machine Learning arXiv:2412.18362 (cs) [Submitted on 24 Dec 2024 (v1), last revised 19 Feb 2026 (this version, v2)] Title:Point-DeepONet: Predicting Nonlinear Fields on Non-Parametric Geometries under Variable Load Conditions Authors:Jangseop Park, Namwoo Kang View a PDF of the paper titled Point-DeepONet: Predicting Nonlinear Fields on Non-Parametric Geometries under Variable Load Conditions, by Jangseop Park and Namwoo Kang View PDF HTML (experimental) Abstract:Nonlinear structural analyses in engineering often require extensive finite element simulations, limiting their applicability in design optimization and real-time control. Conventional deep learning surrogates often struggle with complex, non-parametric three-dimensional (3D) geometries and directionally varying loads. This work presents Point-DeepONet, an operator-learning-based surrogate that integrates PointNet into the DeepONet framework to learn a mapping from non-parametric geometries and variable load conditions to physical response fields. By leveraging PointNet to learn a geometric representation from raw point clouds, our model circumvents the need for manual parameterization. This geometric embedding is then synergistically fused with load conditions within the DeepONet architecture to accurately predict three-dimensional displacement and von Mises stress fields. Trained on a large-scale dataset, Point-DeepONet demonstrates high fidelity, achieving a coefficient of determination (R^2)...
